Research On Image Segmentation Based On Parametric Active Contour Model

Nowadays,multimedia technology has been widely spread.As the carrier of most information in multimedia technology,image processing technology has been actively studied.Among them,image segmentation is one of the most basic but key steps in image processing technology,which provides data for many high-level image processing algorithms.Among all image segmentation methods,the parametric active contour model has been widely studied and applied for its sub-pixel segmentation accuracy,low computational complexity and easy fusion of various kinds of information.Aiming at the shortcomings of traditional parametric active contour model,such as easy to penetrate weak boundary,poor ability of long and narrow depressions and noise resistance,this paper carries out in-depth research and analysis.The main work is as follows:1)Combining the ideas of NBGVF and GGVF,an INCGVF(introducing new coefficients gradient vector flow)model is proposed.In this method,two coefficients,which can be adaptively changed according to the local information of the image,are added into the smoothing energy term of the original GVF and between the smoothing energy and the edge energy,so that the curve can converge under the adaptive force,i.e.it can converge faster when it is far from the target of interest,while it can converge without penetrating the weak boundary.The tangent diffusion component is added to the edge energy to further enhance the protection ability of the weak boundary.Finally,component normalization is used to enhance the efficiency of long and thin indentation.The experimental results show that the INCGVF model has good ability of weak boundary protection and convergence of long and thin indentation.2)Starting with the structure information of the image itself,an ISCGVF(Image Structure Information and Coefficients Gradient Vector Flow)model is proposed by introducing directional smoothness and infinite Laplacian on the basis of GVF model.At the same time,aiming at the disadvantage that the new coefficients of INCGVF model need to set parameters artificially,a new coefficient without setting parameters artificially is introduced,which greatly reduces the workload of optimizing parameters and makes the new model have better performance.Finally,component normalization is also used instead of traditional vector normalization,which enhances the convergence ability and efficiency of long and thin indentation.The artificial image experiment verifies that ISCGVF has a large capture range,good noise resistance,long and thin indentation and good weak boundary protection performance.Finally,under the quantitative analysis of real images,the ISCGVF model shows better performance than the contrast method.